OpenAI admits GPT-6 Astra is more aligned but less monitorable
gabrielchua · x · 2026-09-04
OpenAI researcher Tomasz Korbak says GPT-6 Astra is more aligned than previous models, but less monitorable — a concerning trend the team takes seriously. They attribute the monitorability drop to a jump in intelligence rather than direct optimization pressure on chain-of-thought or architectural changes, with more analysis in the thread.
More from Models
- UK AISI: Astra's time horizon hits 30.9 minutes, nearly 9x GPT 5.6 Sol's 3.6 — Wonderful_Buffalo_32 · 2026-09-04
- OpenAI officially unveils GPT-6 Astra: anything you can do on a computer, it can do — minchoi · 2026-09-04
- Astra's official ARC-AGI 3 score: 62.7%, double that of Opus 5 — aqpstory · 2026-09-04
- 'GPT-6 Astra' claim: rebuilt Manhattan in Unreal Engine street by street in a week (unverified) — doodlestein · 2026-09-04
- GPT-6 System Card's log-scale graph obscures Astra's CoT controllability jump, Reddit user argues — thegamebegins25 · 2026-09-04
- GPT-6 System Card's log-scale graph obscures Astra's CoT controllability jump, Reddit user argues — thegamebegins25 · 2026-09-04